The Methodology

The AI-Assisted
Workflow

Eight steps. From the first conversation with AI to a deployed, documented application. This workflow is the spine of Module 3 and every project that follows.

Module 3 — Workflow overview slide

Screenshot placeholder

The complete workflow at a glance — eight steps from brainstorm to deployed application Module 3, L3.1–L3.8 · The AI-Assisted Workflow overview slide

Brainstorm

Module 3, L3.1

You

Define the problem. Explore the solution space with AI as a thinking partner.

Generates options, challenges assumptions, surfaces edge cases you hadn't considered.

Specify

Module 3, L3.2

You

Write a one-paragraph specification covering task, context, constraints, and output format. Run `git init` before you open the AI tool.

Reviews and improves your specification. Identifies ambiguities before they become rework.

The specification is the bottleneck. Every vague requirement becomes a correction cycle later. A good spec prevents that.

Implement

Module 3, L3.3

You

Direct the AI at the right level (3, 4, or 5). Review each output before moving forward.

Writes code from your specification — functions, files, or multi-file systems depending on the level.

Test

Module 3, L3.4

You

Commit before each agent run. Run tests on the output. Validate against your original specification.

Generates test cases, writes test code, and at Level 5 runs tests and reports results.

Commit before every agent run. An agent that runs unchecked can make plausible-looking changes that break something you weren't watching.

Iterate

Module 3, L3.5

You

Identify what doesn't meet the specification. Reframe the problem, not just the symptom.

Refactors code on request, diagnoses failures, suggests architectural improvements.

Document

Module 3, L3.6

You

Write the README and decision log. Record why, not just what — future-you will be grateful.

Generates docstrings, API docs, and README scaffolding from your codebase.

Deploy

Module 3, L3.7

You

Choose the right deployment target for the audience. Makers deploy differently than enterprises.

Writes deployment configs, Dockerfiles, systemd units, and step-by-step deployment guides.

Maintain

Module 5, L5.6

You

Monitor, update dependencies, and evolve the application as requirements change.

Diagnoses bugs from logs, suggests dependency upgrades, writes migration scripts.

The Core Skill

Four elements of a well-framed prompt

Framing is disciplined communication: giving the AI the information it needs to act correctly. Not all four elements are required for every prompt — but for anything at Level 4, all four should be present.

Task

What should happen? Be specific about the output type.

"Generate a Python function" not "help me with Python"

Context

What does the AI need to know to get it right? Your domain knowledge goes here.

The physics, the hardware, the formula, the expected behaviour

Constraints

What must not happen? Narrow the solution space explicitly.

"No external libraries. Must work offline. Python 3.8+"

Output Format

How should the result be structured? Be explicit about what you want back.

"A single HTML file starting with <!DOCTYPE html>"

L1.3 — Four elements slide

Screenshot placeholder

The four-element framing layout — the visual students photograph and keep for every new project Module 1, L1.3 · The Four Elements of a Well-Framed Prompt · Slide 3

The specification habit

Before you open the AI tool for any task larger than a snippet, write a one-paragraph specification. It doesn't need to be formal. It needs to cover task, context, constraints, and output format. That paragraph is your prompt.

This is a habit, not a technique. The first few times it feels like extra work. After a few sessions it becomes automatic — and the quality and consistency of your outputs will reflect that.

The AI-AssistedWorkflow